How Do You Know if Your Doctor Is Doing a Good Job?

We’ve spent a lot of ink in this blog discussing how difficult it is to measure quality in the various US healthcare systems. One large-scale effort to measure quality is the “Medicare Merit-based Incentive Payment System,” or MIPS. MIPS is a big deal for health systems. Quality isn’t just for professional pride. The MIPS program has a significant impact on the reimbursement received by U.S. physicians.

Some of the surveys or questions you’ve undoubtedly had to answer in doctors’ offices the last few years are undoubtedly tied to their efforts to improve their MIPS score. MIPS rates physicians based on measures in four categories:

1. Quality (30% weight), mostly in terms of clinical outcomes and patient experience. Doctors might be scored on the percentage of hypertensive patients who have their blood pressure controlled or the percentage of their patients who report a high level of satisfaction with their care.

2. Promoting interoperability (25% weight), how well a physician uses technology to improve the quality and efficiency of their care. Measures in this category might include the percentage of patients using the electronic health record (EHR) portal or how many prescriptions are sent to the pharmacy electronically.

3. Improvement activities (15% weight), how well a physician is working to improve her practice through activities like quality improvement programs.

4. Cost (30% weight), how much a physician’s care costs compared to his peers. Think: the number of seemingly unnecessary tests and procedures ordered.

Because the work that, say, a psychiatrist does is so different from the work a urologist does, doctors who participate in MIPS may choose six of a possible 257 performance measures to report, only one of which must be an “outcome measure,” such as hospital admission for a particular illness. The others can be “process measures” like rates of cancer screening. Docs are given a composite MIPS score between zero and 100. To avoid a “negative payment adjustment,” (that is, a reduced fee) physicians must score >75, which seems high to me unless I frame it as a solid “C” grade. Also, 86% of the docs in the sample achieved at least that score, indicating that they either are good at gaming the system or that the score isn’t terribly difficult to achieve.

In spite of the massive effort put into MIPS by regulators, docs, and health systems, it’s unclear whether the MIPS program really reflects the quality of care provided by participating physicians. To investigate, investigators analyzed 3.4 million patients treated in 2019 by 80,246 primary care physicians using Medicare datasets (paywall). They looked specifically at five “process measures” like rates of diabetic eye examinations and breast cancer screens and the “patient outcomes” of all-cause hospitalizations and emergency department visits.

They found that physicians with low MIPS scores (<30) had worse performance on three of the five process measures compared to those with high (>75) MIPS scores. Specifically, the low-scoring docs had lower rates of diabetic eye exams, HbA1c screening for diabetes, and mammography for breast cancer screening. However, the lower-performing docs had better rates of flu vaccination and tobacco screening. In the “patient outcomes,” there was no consistent association with MIPS scores: emergency department visits were lower (e.g., better) for those with low MIPS scores, while all-cause hospitalizations were higher (worse).

Overall, these inconsistent findings suggest that the MIPS program may not be an effective way of measuring and incentivizing quality improvement among U.S. physicians. The “patient outcomes,” which I think most of us would be most interested in, showed no clear association with MIPS scores. In addition, the study found that some physicians with low MIPS scores had very good composite outcomes, while others with high MIPS scores had poor outcomes. Like every correlative study, there were outliers. This suggests that there may be other, more nuanced, factors at play that are not captured by the MIPS program that influence a physician’s performance.

The study is recent enough that we don’t have peer-reviewed criticism or hypothesizing yet about the potential mechanism of MIPS failure. But a blog post from Cornell puts it this way: “…there is inadequate risk adjustment for physicians who care for more medically complex and socially vulnerable patients and that smaller, independent primary care practices have fewer resources to dedicate to quality reporting, leading to low MIPS scores.” So, sicker patients going to smaller, independent practices may drag down results. Put another, more frank, way from Dr. Amy Bond in the same blog post, “MIPS scores may reflect doctors’ ability to keep up with MIPS paperwork more than it reflects their clinical performance.” For our comrades in Human Resources, I suspect this criticism rings especially true.

As the Medical Director of the Kansas Business Group on Health, I’m sometimes asked to weigh in on hot topics that might affect employers or employees. This is a reprint of a blog post from KBGH.

What Health Care Can Learn From Netflix

Streaming video over the internet is a memory hog. If Netflix were to simply store movie files as mp4s and send them out to subscribers on-demand, Stranger Things fans would crash the internet in minutes. Software engineers have solved this problem by creating “compression algorithms” to reduce the file sizes of the transmitted movies. Compression algorithms work by mostly ignoring each frame's composition and instead storing only the changes in the video from frame to frame, so-called “diffs.”

 The transition from one frame to the next is compressible, then, to the extent that it is predictable. Fast movies with a lot of action and cuts, like superhero spectaculars, are hard to compress because of the extra diffs. Slower movies with subtle changes frame-to-frame are easier to compress since the memory required to store and transmit the diffs is small.

 This is analogous to how our own minds work. When we’re left to a single, focused task, we can be remarkably productive. But in the modern workplace, emails, Slack messages, and texts interrupt us more than 150 times a day, and our productivity suffers. Computer engineers call the switch from one task to another a “context switch,” and they don’t like it. Thus, the compression algorithms above. But humans are subject to these context switches, too. Experiments have shown that the average time to recover brain function after a context switch, like interrupting writing this blog post to check an email, is more than 20 minutes. Multitasking is a myth, and most of us spend most of our days in constant recovery from these context switches.

 Now think of how interactions with doctors tend to go. After you’ve traveled 37 minutes traveling to the appointment and spent 64 minutes waiting for her, your doctor enters the room to greet you, often without having reviewed your chart ahead of time. She asks you an open-ended question about how you’re doing, and after a few seconds of pleasantries, you get to your chief complaint for the visit, like your stuffy nose or your back pain or your constipation. The doctor, who is likely trying to type into an electronic health record as you speak, interrupts you after an average of 11 seconds. Then, a nurse knocks on the door to tell your doctor that she has a call from the hospital radiology department on the line. Your doctor leaves the room and comes back a few minutes later, visibly frazzled. You do your best to get the rest of your constipation story out before your doctor sets down her laptop and asks you to climb onto the exam table for an exam. She mostly makes small-talk during the brief exam, then takes a minute to record her findings in the EHR while you wonder if you should peruse the two-year-old copy of People magazine hanging on the wall. You are left to accept the doctor’s recommendations that are based more on pattern recognition and a knowledge of disease epidemiology than any deep thinking about your specific pathology. So she’s wrong about five percent of the time.

 Don’t think of this scenario as a mark against your doctor. Think instead of the system in which she works. How many context switches did your doctor have to navigate to get to the end of your visit? When we point out negative health outcomes in this blog, like the fact that only half of indicated care is delivered to a given patient or that a quarter of care that is delivered may be unnecessary, we’re not out to get doctors. A doctor writes many of these blog posts, and reads all of them. What we’re trying to illuminate are systemic problems.

 Let’s magically teleport you and your doctor into a different system. This time, your doctor has reviewed your chart prior to your visit in a preplanned team “huddle” with her nurses and staff in which your preventive needs have been thoroughly reviewed according to USPSTF guidelines. Your chronic care needs have been addressed mostly outside the clinic visit by periodic communication with a community health worker and a nurse. You’ve sent important biometric information like blood pressures, weights, blood glucose levels, or peak airflow testing, to your doctor’s office already through a secure device or portal. When you get to the clinic, a medical assistant spends twenty minutes with you confirming critical elements of your history, sending predictable refill authorizations to the pharmacy, and predicting changes to your medications based both on the information you’ve sent and on your conversation. Your doctor enters the room knowing that most of your predictable care has been addressed already, and she can confirm or reject the changes in your predictable care that have been “compressed” by the clinical processes in place. Then, she can use most of her brainpower to take care of any unpredictable changes, what the software engineers might call “diffs,” like your new back pain or constipation. And this time, your doctor comes with a medical scribe to take notes for her, so that she doesn’t have to “text and drive” with you in the passenger seat. (in the future, she’ll likely rely on an “ambient” artificial intelligence program to document your visit, but that’s a topic for another day)

Maybe it’s not a surprise that multi-hundred-billion dollar companies get things right sometimes. Netflix has invented a better way for doctors’ offices to function. They just don’t know it.

[disclosure: KBGH receives funding from the Centers for Disease Control and the Kansas Department of Health and Environment to promote team-based care, including community health workers]

As the Medical Director of the Kansas Business Group on Health, I’m sometimes asked to weigh in on hot topics that might affect employers or employees. This is a reprint of a blog post from KBGH.

Do we need to coach patients how to read their notes?

Do you know what is in your medical record? I don’t mean dry lab values or x-ray reports. I mean your doctors’ interpretations of those things, along with intimate, personal details like the results of your physical examinations and their impressions of your adherence to medications and home environment. The information is about you, but it also belongs to you. But it wasn’t always available to you. The 1996 Health Insurance Portability and Accountability Act (HIPAA) gave patients the legal right to review their medical records. But few lay people other than medical malpractice attorneys knew what to do with the information. 

And electronic health record (EHR) vendors have historically treated your data as proprietary, in spite of whatever HIPAA had to say about it. The data was treated as the EHR vendor’s property and was difficult to transfer from one health record to another. Stopgap measures like the Kansas Health Information Network popped up to try to make the data transferable from one hospital or clinic to another. But even this was suboptimal. Compare your experience with your health data to your experience with your financial data, which is probably almost as sensitive. You have undoubtedly used your ATM card from, say, Intrust Bank, to check your balance at, for example, a Fidelity ATM. We take it for granted, just like I take for granted that my USB drive from my home Mac computer will plug into my work PC. I don’t have to rely on the substantial expertise of a middle man to know that the data will transfer.

Since the 1990s several experiments have led to a movement for patients to have ready access to their doctors’ notes, be they on paper or in an electronic format. The best known organization goes by the name “Open Notes.” Now, new federal rules stemming from the 21st Century Cures Act aim to promote further patient access to their electronic health records via secure online “portals.” With a few exceptions, starting April 5, 2021, clinical notes and much other electronic information must be made available free of charge to patients. And the new rule forbids health systems or electronic health record vendors from “information blocking,” the practice of treating electronic health data as a proprietary asset and restricting access. The Annals of Internal Medicine (paywall) has a nice infographic:

Annals of Internal Medicine

To help folks transition to this new world, I think employers, insurers, and health care providers need to be proactive. To start, we should encourage our patients or employees to find their health record and to discuss it with their treating practitioner. Medical records are teeming with mistakes, due to cut-and-paste, poor user interoperability, and old-fashioned errors. Patients who find these mistakes shouldn’t run straight to the nearest malpractice attorney. Instead, both the patient and the doctor might be enlightened by a discussion of what should have been in the note.

Second, we should teach our employees and patients that what is in the record is meant to be objective. It is not a subjective judgement of anyone’s value as a human being. When I was in full-time academic practice, I remember a colleague being berated by a patient for noting that the patient smelled like urine (because she did, objectively, smell like urine). But once the patient’s embarrassment over reading the note faded, she was able to have a very meaningful conversation with her doctor about her urinary incontinence, and she was prescribed a medication that helped tremendously. Without access to her record this may never have happened. 

Finally, we need to remind patients that their record is truly private. In the era of predatory social media companies we’ve largely given up on the idea that any of our personal information should be private. But Facebook is probably not the best forum in which to litigate disagreements with doctors or nurses or to share screenshots of one’s medical record.

As the Medical Director of the Kansas Business Group on Health I’m sometimes asked to weigh in on hot topics that might affect employers or employees. This is a reprint of a blog post from KBGH.